Microsoft unveiled Project Perception, a cybersecurity platform that orchestrates AI models from Microsoft, OpenAI, and Anthropic simultaneously. The system enters public preview on August 3 and uses intelligent routing to assign each task to the most suitable model. On the CyberGym benchmark, the underlying MDASH model scored 88.45% in May 2026, surpassing Anthropic’s Mythos (83.1%) and OpenAI’s GPT-5.5 (81.8%). Microsoft positions this solution as a lower-cost alternative, arguing that routing to the most efficient model reduces overall compute spend. The announcement comes in response to Anthropic’s cybersecurity initiative (Project Glasswing) and recent changes to OpenAI’s Trusted Access for Cyber program.
Source: Read the original article
